三维激光扫描点云数据分类去噪及空洞修复算法研究

发布时间：2018-04-11 00:34

本文选题：三维激光扫描 + 点云分类去噪　；参考：《长安大学》2017年硕士论文

【摘要】：三维激光扫描技术作为一种全新的测绘技术,采用非接触式的方式快速获取被测物体三维坐标信息。受偶然误差和系统误差的影响,三维激光扫描设备采集的点云数据中不可避免的掺杂着一些噪声数据,这些噪声数据如果不及时进行去除,不但会增加数据量,而且影响后期建模的精度以及数据处理效率;三维激光扫描点云数据噪声点通常分为两种,即远离点云主体的大尺度噪声和靠近点云主体、混杂在被测物体模型表面的小尺度噪声。目前所采用的大多数去噪算法都是利用相同的处理方法实现对不同类型的点云数据去噪,虽然能够实现对噪声点的去除,但是也会因此删除很多属于模型本身的尖锐特征造成过光顺问题,导致被测物体建模后失真。本文针对这种问题,提出针对不同类型点云噪声数据的分类去噪算法,实现点云数据的高效去噪。本文主要研究内容如下:(1)在点云去噪过程中,充分利用PCL点云库处理点云数据的优势,基于PCL中集成的点云滤波模块进行点云数据分类去噪。首先利用Statistical Outlier Removal滤波器、Radius Outlier Removal滤波器实现大尺度噪声的去除,然后通过改进BilateralFilter类中的双边滤波算法实现小尺度点云噪声的去除。最后应用具体的工程实例验证该分类去噪算法的可行性。(2)针对存在空洞的点云数据,利用RBF神经网络对需要处理的点云数据进行预测模拟,将点云实测值全部用网络预测值代替,实现点云数据的去噪。通过实例验证得出,该算法能够有效去除点云表面的噪声,保持点云模型的特征信息,在点云去噪过程中实现空洞修复,提高了点云数据处理的效率。
[Abstract]:As a new technology of surveying and mapping, 3D laser scanning technology acquires 3D coordinate information of measured object quickly by non-contact method.Under the influence of accidental error and systematic error, the point cloud data collected by 3D laser scanning equipment is inevitably mixed with some noise data. If the noise data is not removed in time, it will not only increase the amount of data,The noise points of 3D laser scanning point cloud data are usually divided into two types: the large scale noise away from the point cloud body and the near point cloud subject.Small scale noise mixed on the surface of the measured object model.Most of the current de-noising algorithms use the same processing method to realize the de-noising of different types of point cloud data, although the noise points can be removed.However, many sharp features which belong to the model itself will be deleted, which will lead to over-fairing, resulting in distortion after modeling of the object under test.In order to solve this problem, this paper proposes a classification denoising algorithm for different types of point cloud noise data to realize the efficient denoising of point cloud data.The main contents of this paper are as follows: (1) in the process of point cloud denoising, we make full use of the advantage of PCL point cloud database to deal with point cloud data, and carry out point cloud data classification and denoising based on the integrated point cloud filtering module in PCL.Firstly, the Statistical Outlier Removal filter Radius Outlier Removal filter is used to remove large scale noise, and then the small scale point cloud noise is removed by improving the bilateral filtering algorithm of BilateralFilter class.Finally, a concrete engineering example is used to verify the feasibility of the classification denoising algorithm. (2) aiming at the point cloud data with holes, the RBF neural network is used to predict and simulate the point cloud data to be processed.The point cloud data denoising is realized by replacing the point cloud data with the network prediction value.The results show that the algorithm can effectively remove the noise on the surface of the point cloud, keep the characteristic information of the point cloud model, repair the void in the process of point cloud denoising, and improve the efficiency of point cloud data processing.
【学位授予单位】：长安大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：P225.2

【相似文献】